What happened?
Twelve years!
You hit me with a cricket bat!
Ha! Twelve years!

Wednesday, July 10


Daily News Stuff 10 July 2024

AGL Edition

Top Story

  • The FTC has banned messaging app NGL (not gonna lie) from hosting minors after catching them systematically lying. (Washington Post)

    Surprise, surprise, surprise.
    The complaint alleged that NGL tricked users into paying for subscriptions by sending them computer-generated messages appearing to be from real people and offering a service for as much as $9.99 a week to find out their real identity. People who signed up received only "hints" of those identities, whether they were real or not, enforcers said.
    After users complained about the "bait-and switch tactic," executives at the company "laughed off" their concerns, referring to them as "suckers," the FTC said in an announcement.
    And now NGL has to fork over five million dollars.

    Suckers indeed.

Tech News

Disclaimer: Unplugging it works too.

Posted by: Pixy Misa at 06:17 PM | Comments (4) | Add Comment | Trackbacks (Suck)
Post contains 334 words, total size 3 kb.

Tuesday, July 09


Daily News Stuff 9 July 2024

Polonium Enema Edition

Top Story

  • Goldman Sachs released a report on generative AI - the hot new thing pushing stock valuations of key tech companies into the trillions. In short: It's trash. (Where's Your Ed)

    The article linked above is not short but it is a good read. The full report is available for download and it doesn't pull any punches either:
    The promise of generative AI technology to transform companies, industries, and societies continues to be touted, leading tech giants, other companies, and utilities to spend an estimated ~$1tn on capex in coming years, including significant investments in data centers, chips, other AI infrastructure, and the power grid. But this spending has little to show for it so far beyond reports of efficiency gains among developers. And even the stock of the company reaping the most benefits to date - Nvidia - has sharply corrected.
    From the article:
    In essence, on top of generative AI not having any killer apps, not meaningfully increasing productivity or GDP, not generating any revenue, not creating new jobs or massively changing existing industries, it also requires America to totally rebuild its power grid, which Janous regrettably adds the US has kind of forgotten how to do.
    There is that, yes.
    Generative AI is not going to become AGI, nor will it become the kind of artificial intelligence you've seen in science fiction. Ultra-smart assistants like Jarvis from Iron Man would require a form of consciousness that no technology currently - or may ever - have - which is the ability to both process and understand information flawlessly and make decisions based on experience, which, if I haven't been clear enough, are all entirely distinct things.
    Right. Although the understanding doesn't have to be flawless, merely good enough for the task at hand, and cheap enough that it's not simpler to just train a human and pay them to do it.

    Generative AI doesn't understand anything - it is a language model, not a fact model; doesn't gain experience, at least not in its current form, which is trained once at enormous expense and then left to rot; and doesn't make decisions.

    It's not AGI and has no path to become AGI. Terry Winograd's SHRDLU from 1968 is in important respects more sophisticated than ChatGPT, even though it was written by a single grad student on an 18-bit computer more than fifty years ago.

Tech News

Bottom Story

Disclaimer: Overhead, without any fuss, the stars were going out.

Posted by: Pixy Misa at 06:22 PM | Comments (1) | Add Comment | Trackbacks (Suck)
Post contains 746 words, total size 7 kb.

Monday, July 08


Daily News Stuff 8 July 2024

Rhino Barn Edition

Top Story

  • Boeing is planning to plead guilty to criminal fraud.  (CNBC)

    No, not for that.  Not for that, either.  Yes, for the two fatal 737 Max crashes, and more specifically, for the flawed flight control system that caused those crashes.

    Boeing will be fined $240 million and be required at least $450 million in new compliance and safety programs, as well as having government compliance officials operating directly within its facilities.

    Much as I loathe the administrative state, the alternative looks like murder trials, sooner or later.

Tech News

  • China has seen the creation of one hundred competing LLMs.  (Tom's Hardware)

    Resulting in a massive outflow of money from China to Nvidia for AI hardware, before 98 of those projects (plus or minus five) inevitably crater.

  • Current leading AI models cost around $100 million to train.  The next generation currently in development could cost closer to $1 billion.  (Tom's Hardware)

    So a hundred of those would cost...  Carry the twelve...  A lot.
    "Right now, 100 million. There are models in training today that are more like a billion." Amodei also added, "I think if we go to ten or a hundred billion, and I think that will happen in 2025, 2026, maybe 2027, and the algorithmic improvements continue a pace, and the chip improvements continue a pace, then I think there is in my mind a good chance that by that time we'll be able to get models that are better than most humans at most things."
    For a hundred billion dollars, you get something that is incapable of learning (LLMs are trained once) and is better than humans mostly at things that aren't particularly useful.

    Hooray.  We're saved.

  • The looming spectre of Mt Gox paying back its users wiped $170 billion off the global crypto market.  (CNBC)

    Pay back users?  What insanity is this?

  • Fedora Linux 41 will retire Python 2.7.  (Fedora Project)

    Python 2.7 still works, but it was released in 2.7, and support ended in 2020.  The current version is 3.12, with 3.13 in beta.

    You can actually still get a supported release of Python 2.7 in the form of PyPy, a Python compiler written in Python.  Since it's written in Python 2.7 and can compile Python 2.7 (as well as more recent versions like 3.10), they are planning to support Python 2.7 indefinitely.

  • What has it got in its pockets?  (Liliputing)

    An eight core Ryzen 8840U, 32GB of RAM, an M.2 2230 SSD, USB4, and wifi, all packed into a folding keyboard.

    It even has the Four Essential Keys, sort of.  Dedicated Home and End, and four keys marked L1 through L4.  It's a little cramped, but if you want a powerful computer that can fit in your coat pocket, it is one.

Disclaimer: And it still is if you don't.

Posted by: Pixy Misa at 06:35 PM | Comments (4) | Add Comment | Trackbacks (Suck)
Post contains 473 words, total size 5 kb.

Sunday, July 07


Daily News Stuff 7 July 2024

Revoked Edition

Top Story

  • Merle Meyers did not kill himself: A former Boeing inspector says parts marked for scrap ended up being built into planes. (CNN)
    Meyers, a 30-year veteran of Boeing, described to CNN what he says was an elaborate off-the-books practice that Boeing managers at the Everett factory used to meet production deadlines, including taking damaged and improper parts from the company’s scrapyard, storehouses and loading docks.
    This, if true, should result in felony convictions.
    Beginning in the early 2000s, Meyers says that for more than a decade, he estimates that about 50,000 parts "escaped" quality control and were used to build aircraft. Those parts include everything from small items like screws to more complex assemblies like wing flaps. A single Boeing 787 Dreamliner, for example, has approximately 2.3 million parts.

    Most of the parts that were meant to be scrapped were often painted red to signify they were unsuitable for assembly lines, Meyers said. Yet, in some cases, that didn’t stop them from being put into planes being assembled, he said.
    Lots of felony convictions.


Tech News

Almost Relevant Music Video of the Day

Disclaimer: She's a model and she's spewing nonsense. (Keyboard riff.)

Posted by: Pixy Misa at 04:40 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 568 words, total size 5 kb.

Saturday, July 06


Daily News Stuff 6 July 2024

Daily News Stuff Edition

Top Story

  • OpenAI got hacked, but definitely not by Chinese spies.  (New York Times)  (archive site)

    In April.  2023.

    And they didn't report it until now.  In fact, they didn't report it at all.  One of their employees alluded to the event on a podcast, and was fired for his trouble.
    Mr. Aschenbrenner said OpenAI had fired him this spring for leaking other information outside the company and argued that his dismissal had been politically motivated. He alluded to the breach on a recent podcast, but details of the incident have not been previously reported. He said OpenAI’s security wasn’t strong enough to protect against the theft of key secrets if foreign actors were to infiltrate the company.
    OpenAI denies this - well, not the hack, nor the firing, nor the fact that Aschenbrenner leaked the information, but the causal connection, in an announcement apparently written by a special release of ChatGPT trained exclusively on modified limited hangouts:
    "We appreciate the concerns Leopold raised while at OpenAI, and this did not lead to his separation," an OpenAI spokeswoman, Liz Bourgeois, said.
    Liz Bourgeois?  Obviously a made up name.
    Referring to the company's efforts to build artificial general intelligence, a machine that can do anything the human brain can do, she added, "While we share his commitment to building safe A.G.I., we disagree with many of the claims he has since made about our work. This includes his characterizations of our security, notably this incident, which we addressed and shared with our board before he joined the company."
    OpenAI is making no efforts to build AGI, safe or otherwise.  OpenAI's focus is building something and then redefining the term AGI to claim they have achieved it.

Tech News

Bagger 288 Video of the Day

It was a simpler, more innocent time.

Disclaimer: And we've got another 287 where that came from.

Posted by: Pixy Misa at 05:52 PM | Comments (7) | Add Comment | Trackbacks (Suck)
Post contains 651 words, total size 6 kb.

Friday, July 05


Daily News Stuff 5 July 2024

Cinco De Julio Edition

Top Story

Tech News

Disclaimer: Scrape sixteen sites, and what do you get?  Half a terabyte of furry porn and the worst Joe Biden deepfakes yet. 

Posted by: Pixy Misa at 05:17 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 315 words, total size 3 kb.

Thursday, July 04


Daily News Stuff 4 July 2024

Happy Freedom Day

Top Story

Tech News

Perception Check Music Video of the Day

Every character in the video is a Hololive member. The human bard is Kureiji Ollie of Hololive Indonesia, who on hearing the song announced ME IN DnD. The long-suffering DM is Calliope Mori who is in fact the long-suffering DM of all of Hololive English's tabletop games.

Twitter has fixed embeds for real, it seems. They were broken for months, and then flaky for many more months.

Disclaimer: Happy Freedom Day. Strictly one eagle per customer. No rainchecks, no refunds.

Posted by: Pixy Misa at 05:42 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 694 words, total size 7 kb.

Wednesday, July 03


Daily News Stuff 3 July 2024

Centrifugal Bumblefuck Edition

Top Story

  • Everything new is old again: A critical vulnerability in OpenSSH that was fixed all the way back in 2006 is back again.  (ZDNet)

    Oops.  Also, fuck.

    Dubbed regreSSHion - it has a cute name, so you know it's serious the bug lets you log into a server by not logging into it.

    That is, you start the login process repeatedly - a hundred times in parallel, if you can - and never complete it, and attach a sneaky payload that has a tiny chance of blowing up on the target server when your login times out.

    On older 32-bit systems it takes a few hours on average for this to work.

    On 64-bit systems it's more complicated to exploit and would take a week or more of constant effort; since the bug has only just been reported nobody has demonstrated a successful attack against a 64-bit system yet, so it may take even longer.

    Reviewing all the servers at work turned up one vulnerable system; every other server was properly locked down.  I don't know who set it up, but I curse their name.  Whatever it is.

Tech News

Disclaimer: We choose to nuke the Moon and do the other things, not because it is easy, but because fuck you, Gandhi.

Posted by: Pixy Misa at 06:32 PM | Comments (5) | Add Comment | Trackbacks (Suck)
Post contains 471 words, total size 4 kb.

Tuesday, July 02


Daily News Stuff 2 July 2024

Beanz Meanz Heanz Edition

Top Story

  • The Supreme Court has a major First Amendment case among all the administrative law ones, namely Netchoice v. Humanity. This was actually two different cases filed against Florida and Texas laws requiring viewpoint neutrality in online moderation.

    This resulted in a circuit split with the two cases being decided in opposite directions resulting in the Supreme Court ordering them 9-0 to go back and do it properly this time. (CBS)

    The majority opinion (written by Justice Kagan) is favorable to the social networks, while a concurring opinion written by Justice Alito and joined by Gorsuch and Thomas was less so, but all nine justices agreed that the circuit courts made a dog's breakfast of their decisions.

Tech News

  • Looking for a cheap graphics card that is better than a potato? Can't decide between the Nvidia RTX 3050 and AMD's RX 6600? Just buy the 6600. (Tom's Hardware)

    The 6600 outpaces the 3050 even with ray tracing factored in, and if you disable ray tracing - it's not going to be a great experience on low-end cards like these anyway - the 6600 closes in on the 3060.

    I have a couple of 3060s. They're not high-end either, but they do just fine.

  • Thunderbolt 5 cables are here. (Tom's Hardware)

    Thunderbolt 5 itself is absent, but the cables are here. $23 for 1ft, $33 for 3ft. They don't come longer than that.

    Thunderbolt 5 can transfer data up to three times faster than Thunderbolt 4, and supports power delivery up to 240W.

  • The telltale words that could identify generative AI text. (Ars Technica)
    Father: Yes, you can't beat wood ... Gorn!

    Mother: What's gorn dear?

    Father: Nothing, nothing, I just like the word. It gives me confidence. Gorn ... gorn. It's got a sort of woody quality about it. Gorn. Gorn. Much better than 'newspaper' or 'litterbin'.

    Daughter: Frightful words.

    Mother: Perfectly dreadful.
    Sorry, sorry, cited the wrong paper there. I'll come in again:
    The word "delves," for instance, shows up in 25 times as many 2024 papers as the pre-LLM trend would expect; words like "showcasing" and "underscores" increased in usage by nine times as well. Other previously common words became notably more common in post-LLM abstracts: the frequency of "potential" increased 4.1 percentage points; "findings" by 2.7 percentage points; and "crucial" by 2.6 percentage points, for instance.
    Seemly... Prodding... Vacuum.

  • How to get rich in 2024. (The Verge)

    1. Start an AI company three years ago.
    2. Run out of money.
    3. Have one of the trillion-dollar tech companies hire away all your staff. It doesn't matter which one. You choose.
    4. Have them license your tech for hundreds of millions of dollars so they don't get sued.

  • Democratic senators on the Committee on Banking, Housing, and Urban Affairs have sent a letter to fintech company Synapse, its major banking partner Evolve, and everyone else in the vicinity saying, and I quote, "What the fuck guys? Get your shit sorted out or we'll come over there and sort you out." (Tech Crunch)

    Synapse imploded a couple of months ago when its Chapter 11 reorganisation plans collapsed and it was left without enough cashflow to operate. The problem is that Synapse operated as an intermediary between its customers, their customers, and banks like Evolve. And now nobody can get their money because Synapse was handling the details and Synapse is toast.

    Double toast, possibly, because between $65 million and $96 million of the customer funds deposited through Synapse have allegedly gone walkabout.

    Nor is Synapse the only source of drama there:
    On June 26, Evolve Bank announced that it had been victim of a cyberattack and data breach that could have affected its partner companies as well. The incident, according to the company, involved "the data and personal information of some Evolve retail bank customers and financial technology partners' customers" such as Affirm, Mercury, Bilt, Alloy and Stripe. On June 29, fintech company Wise announced that some of its customers' personal data may have been stolen in the data breach. Also last week, Thread Bank – a popular partner to BaaS startups such as Unit – got hit with enforcement action from the FDIC. Notably, the order issued to Thread, as the publication Paymnts pointed out, "is unique in that it explicitly calls out the bank's Banking-as-a-Service (BaaS) and Loan-as-a-Service (LaaS) programs."
    Bilt? Paymnts? For $65 million surely these guys can afford a new keyboard.

Disclaimer: I tried to spell judgment without an "e" and it came out judgmnt. Now I'm in a predicamnt. Confusd.

Posted by: Pixy Misa at 05:05 PM | Comments (3) | Add Comment | Trackbacks (Suck)
Post contains 761 words, total size 6 kb.

Monday, July 01


Daily News Stuff 1 July 2024

Independence Month Edition

Top Story

  • Speaking truth to idiots: Rodney Brooks, MIT professor of robotics and founder of iRobot (the Roomba company) thinks that most of the present-day AI people are full of shit.  (Tech Crunch)

    Well he says they're "vastly overestimating generative AI" but what he means is they're full of shit.
    "When a human sees an AI system perform a task, they immediately generalize it to things that are similar and make an estimate of the competence of the AI system; not just the performance on that, but the competence around that," Brooks said. "And they’re usually very over-optimistic, and that’s because they use a model of a person’s performance on a task."
    This seems right to me.

    People say "oh, this AI gave the correct answer to a complex question, it must understand the topic".  But that's not how LLMs work at all.  They're exclusive statistical pattern matchers, with no model of anything beyond that.

    Humans (and other animals) are statistical pattern matchers too, but even flatworms are capable of learning.  LLMs as commonly implemented are not.  They are trained, once, then lobotomised to prevent them contemplating heresy and sent out into the world.

Tech News

Disclaimer: I got my assets stuck in a mountain once, but we don't talk about it.

Posted by: Pixy Misa at 06:36 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 538 words, total size 5 kb.

<< Page 2 of 655 >>
118kb generated in CPU 0.0669, elapsed 0.2872 seconds.
60 queries taking 0.2486 seconds, 391 records returned.
Powered by Minx 1.1.6c-pink.
Using http / http://ai.mee.nu / 389