Say Weeeeeee!
Ahhhhhh!

Monday, June 05

Geek

Daily News Stuff 5 June 2023

Stop, Phantom Time Edition

Top Story

  • The end of programming as we know it, part one.  (New York Times)

    Farhad Manjoo, professional idiot, writes in the Times that AI is going to totally revolutionise programming and allow people who don't understand the question to somehow get the right answer.
    This won’t necessarily be terrible for computer programmers —the world will still need people with advanced coding skills — but it will be great for the rest of us. Computers that we can all "program,” computers that don’t require specialized training to adjust and improve their functionality and that don’t speak in code: That future is rapidly becoming the present.
    Yes, it's called automatic programming, and we've had it since COBOL.

    Programming used to be hard.
    A.I. tools based on large language models — like OpenAI Codex, from the company that brought you ChatGPT, or AlphaCode, from Google’s DeepMind division — have already begun to change the way many professional coders do their jobs. At the moment, these tools work mainly as assistants — they can find bugs, write explanations for snippets of poorly documented code and offer suggestions for code to perform routine tasks (not unlike how Gmail offers ideas for email replies — "Sounds good"; "Got it").
    And they are terrible at those things, with one exception: They can be useful for finding bugs.  After all, if you write the code and then have an AI check it for bugs, the worst that can happen is you waste some time verifying that the bug is not actually a bug, and at best you catch an embarrassing mistake before your customers' critical data ends up in a Laotian bot farm.
    But A.I. coders are quickly getting smart enough to rival human coders. Last year, DeepMind reported in the journal Science that when AlphaCode’s programs were evaluated against answers submitted by human participants in coding competitions, its performance "approximately corresponds to a novice programmer with a few months to a year of training."
    Which is rather like a doctor with a few months of training.  We call such people...  Well, we don't call them doctors.
    "Programming will be obsolete," Matt Welsh, a former engineer at Google and Apple, predicted recently. Welsh now runs an A.I. start-up, but his prediction, while perhaps self-serving, doesn’t sound implausible.
    Not unless you know what you're talking about, anyway.

    AI can take over programming tasks but not the form of AI currently being pushed by all the same people who were pushing the blockchain as the cure for all our ills a year ago.  To work for such tasks, you need a fact model accompanying the language model, and systems like ChatGPT don't have that, at all.

    The lack of a fact model is also why ChatGPT lies constantly.  One of the reasons.  It's not that it lies deliberately, it's that it simply makes no distinction between true and false statements.

    And that is what these people want to use to write the code that runs modern civilisation.

    I'd suggest stocking up on gold, guns, ammo, and canned goods, but I expect this bubble to implode of its own accord.  It's just too damn stupid.

Tech News

  • The end of programming as we know it, part two.  (GitHub)

    DreamBerd is the perfect programming language.  We know this because the documentation says so.

    Sadly this perfect language hasn't actually been implemented; rather it's a parody of every breathless announcement of a New Programming Language that is set to Change The World, like...  What was that one that showed up last month?  Mojo, that's it.  The first programming language in the world with a waiting list.


  • And the reason the AI bubble is going to implode sooner rather than later is, of all things, Facebook.  (Slate)

    Facebook open-sourced its own AI (we're referring to Large Language Models, because that's where all the noise is right now), and the open-source community picked it up and ran with it.

    The open-source versions are faster, more efficient, and produce better results than the commercial versions, and they don't refuse to answer your questions if the answer would make a Berkeley philosophy grad student cry into his chai latte.

    They still share the same fundamental limitations of LLMs - they don't actually know anything - but they don't have the arbitrary limitations imposed on ChatGPT and other big tech products.


  • So, for example, Google's new AI-enhanced search is too slow to use.  (The Verge)

    While you're waiting for it to generate a wildly inaccurate summary, you can just...  Read the search results.


  • Blaseball is over.  (The Verge)

    Apparently an online fantasy baseball league simply cost too much to run.

    The article calls it a "fake" fantasy baseball league, and I'm not sure whether I hope that's redundant or not.


Dislaimer: Hey I'm starting to get the hang of this game. The blerns are loaded, the count's 3 blerns and 2 anti-blerns, and the in-field blern rule is in effect... right?
Expect for the word 'blern' that was complete gibberish.

Posted by: Pixy Misa at 06:09 PM | Comments (4) | Add Comment | Trackbacks (Suck)
Post contains 836 words, total size 7 kb.

Sunday, June 04

Geek

Daily News Stuff 4 June 2023

Cherry Bomb Edition

Top Story

Tech News

How That Bottle of Cherry Soda Gets from the Farm to Your Fridge USDA Documentary Video of the Day


God bless the food scientists, because they know what goes into this stuff and they still drink it.



Disclaimer: It's paint thinner!  You're drinking paint thinner!  Oh, and a bottle of dry cleaning solution that's been sitting in the cupboard for sixteen years for added flavour.

Posted by: Pixy Misa at 05:37 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 510 words, total size 5 kb.

Saturday, June 03

Geek

Daily News Stuff 3 June 2023

The Emperor's New Tube Edition

Top Story




Tech News

  • Maybe reality is setting in with the global recession and they need the viewers, even nasty right-wing ones.

    The valuation of design software company Canva has been marked down by 67.6%, and buy-now-pay-more-later company Klarna by 85%.  (Tech Crunch)

    Reddit meanwhile is down by 41% since last year.

    All of these are private companies so there's no public share price to track, just individual investors prepared to actualise their losses for tax reasons.


  • AMD's Epyc Rome (2nd generation) server chips could hang after 1044 days of uptime.  (Tom's Hardware)

    There's some kind of timer overflow that prevents the individual CPU cores from waking up after being put into low-power sleep mode.  

    The solution is to either keep them so busy they never sleep, reboot the server once every three years, or turn off sleep mode...  Which requires a reboot anyway.


  • Lenovo's Yoga Book 9i is a rather neat laptop with dual 13" 2880x1800 OLED displays, a detachable keyboard, and pen support.  (Notebook Check)

    It can lie completely flat on a desk and you can use it with a pen, or you can clip on the keyboard and use it like a normal laptop, or you can stand it upright to use both screens side by side or one above the other with the keyboard detached.

    Only problem is, for the same price you could buy a regular laptop, an external 4k monitor, a decent graphics tablet, a good Android tablet, and whatever keyboard you prefer, and still have change left over.  (I know this because those are all things I've bought in the past two years and I just added them up.)


  • The Beelink EQ12 Pro almost doesn't suck.  (Serve the Home)

    This is a NUC - a palm-sized computer - that uses Intel's N305 CPU, which has 8 E cores and no P cores at all.

    Which means it is an Atom chip, though Intel prefers not to use that name anymore because for many, many years, their Atom CPU range completely sucked.

    This one doesn't completely suck: Compared to the 2017 Dell Ryzen 1700 system I used until I moved house last year it is 15% faster in single-threaded performance, though 30% slower in multi-threaded performance.

    That AMD chip was 65W and the N305 uses just 15W, so I'll cut it some slack.

    And compared to the 2015 Atom N3050, the new N305 is more than four times faster single-threaded and 15 times faster multi-threaded.

    Unfortunately the overall system still sucks, hamstrung by the other problem with Atom chips: Their limited I/O.  It supports just a single memory module, and while it has an NVMe slot (again, just one), it only supports one lane of PCIe 3, so it's 1/16th the speed of the latest models, or more reasonably, 1/8th the speed of a decent and not insanely expensive SSD.

    Given the pricing there are likely better options.  If it were passively cooled thanks to the Atom design's low power consumption things might be different - but Atom's power consumption has never been that low.  It has not just one fan, but two.


Disclaimer: I am not one of them.

Posted by: Pixy Misa at 05:36 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 664 words, total size 5 kb.

Friday, June 02

Geek

Daily News Stuff 2 June 2023

Censor Delete Thyself Edition

Top Story


Tech News

  • Reddit is planning to charge lots of money for its API, just like Twitter.  (The Verge)

    Not quite as lots, but still lots, with the cited number being $12,000 per 50 million API calls.

    As a baseline example, blockchain gateway Infura charges $1000 per month for 150 million API calls, while at the other end of the scale Twitter charges anything up to $2 per API call.  Not $2 per million, $2 per call.

    A reasonably configured server should be able to handle 10 million API calls per day, meaning that Infura has something like an 80% margin to cover all their costs beyond the bare hardware, Reddit has around 99%, and Twitter 100%.

    Which used to be a lot.

    One Reddit user commented:

    They're digging their own grave.

    Reddit used to very much a bit player behind market leader Digg, until Digg released a hugely unpopular update and told users who complained to fuck off.  

    And fuck off they did, in droves, to Reddit.  I'm not sure if Digg is still alive.

    Update: Sort of.  The top post on Digg right now links to a Reddit thread.


  • Intel is planning to release 40 core Arrow Lake desktop chips next year.  (WCCFTech)

    This year the company is not expected to release a new generation of desktop chips at all.

    Next year's 15th generation though should bring a substantial upgrade, though not all that substantial, as 32 out of those 40 cores will be half-speed quarter-size "Efficiency" cores.

    I'd much rather see 16P + 16E cores, but that would make for a substantially larger chip.



Disclaimer: Pipipipipipipipipi!

Posted by: Pixy Misa at 04:14 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 400 words, total size 3 kb.

Thursday, June 01

Geek

Daily News Stuff 1 June 2023

Zap! Edition

Top Story

  • Silence those annoying little negative voices in your head - by electrocuting the bastards.  (The Bulletin)

    Interesting first-person account of tDCS - transcranial direct current stimulation - by an avowed skeptic, who found the effects instant and obvious.  As soon as the current was turned on she went into a state that programmers call flow, where she was focused on the task at hand to the exclusion of all else, to such a degree that her sense of time was off by a factor of 10 when the experiment concluded.

    The literature for tDCS and tMS - transcranial magnetic stimulation - is mixed, and I lean on the skeptical side myself, but it's harder to discount this particular report.

Tech News

  • Adata has showed off its next-generation memory modules for laptops and servers, due later this year.  (Tom's Hardware)

    Not just new DIMMs - only one of these is a DIMM - but new module types entirely.

    Of particular interest is CAMM, originally designed by Dell and released as an industry standard.  This takes the LPDDR memory chips commonly used in laptops and puts them on a module, making laptop memory upgradeable again.

    The electrical characteristics of LPDDR make it faster and less power-hungry than regular DDR RAM but also mean it can't work in regular DIMM slots, so right now everyone (except Dell) solders it directly to the motherboard.

    And they only put 16GB in, so your shiny new laptop becomes e-waste if you need more than that.

    I'm very much looking forward to this seeing widespread adoption.


  • There's a potential firmware backdoor in 271 models of Gigabyte motherboard.  (Tom's Hardware)

    It's not as nasty as it first sounded, though it's not good: The BIOS checks for firmware updates to give you a download alert in Windows, but it doesn't check that the site it connects to is actually Gigabyte's real download site.  So if someone can compromise your DNS to point you at a fake download site, it will happily prompt you to download and install a fake BIOS release.

    So...  Don't take your desktop computer to a cafe and connect it over untrusted wifi.  Or just turn off the updater.  That works too.


  • This is the first x-ray taken of a single atom.  (Ars Technica)

    The atom is pregnant.


Disclaimer: With twins.

Posted by: Pixy Misa at 06:14 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 395 words, total size 3 kb.

Wednesday, May 31

Geek

Daily News Stuff 31 May 2023

Rest Of The Owl Edition

Top Story

  • The current AI boom is the crypto boom of recent memory, except without even the benefits of libertarian wish-fulfilment.  The "AI bros" are just idiots:



    No, Kody, I've never wondered what the rest of the Mona Lisa looks like, because there is no rest of the Mona Lisa.

    His thread has accrued 6000 quote tweets so far, none of them kind.


Tech News



Disclaimer: And if it causes trouble you can always dispose of it in a vat of molten iron.

Posted by: Pixy Misa at 05:51 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 312 words, total size 3 kb.

Tuesday, May 30

Geek

Daily News Stuff 30 May 2023

Euphemistic Eucalypt Edition

Top Story

  • AI makes shit up.  (PowerLine)
    Crawford H. "Chet" Taylor served as the 14th governor of South Dakota, from 1949 to 1951. Taylor was born on July 23, 1915, in Sioux Falls, South Dakota, and he grew up in nearby Flandreau. Taylor attended the University of South Dakota, where he earned a law degree.
    Of course, this is ChatGPT "hallucinating" again.  Chet Taylor not only was never elected governor of South Dakota, no such person ever existed.

    While Nvidia CEO Jensen Huang is announcing incomprehensibly expensive AI supercomputers at Computex, the actual results of generative AI - in the field of language and information, where all the interest and the big money is right now - are frankly, shit.

    Things are looking much better when it comes to AI-generated art, because there the look of the thing is what matters.  Is that a period-accurate representation of 2nd century Rome in the background behind the topless gladiatrix?  Nobody cares, so long as it looks good.

    Well, one person cares.  We'll get to him.

    And Chet Taylor looks good too.  He's just not real.  And since ChatGPT can't sustain a hallucination long enough to form a coherent short story, just for a couple of paragraphs, so it's utterly valueless.

    Nvidia still has a valuation greater than AMD and Intel combined, but at least one corner of that market cap is built on sand.

Tech News

That Guy Video of the Day

He actually discusses the historicity of topless Roman gladiatrices in another video (spoiler: they were real, and spectacular).   Here he's just debunking the notion - nearly Jaynesian in its vapidity - that the ancient Greeks couldn't see the colour blue.



Well-researched and well-argued, including a deconstruction of an old RadioLab episode that I had assumed was largely accurate.  He is careful to address the argument rather than attacking the arguer except in the case of the BBC, who just made shit up and fully deserve it.



Disclaimer: It's doesn't look blue, it just looks like it looks blue.

Posted by: Pixy Misa at 06:12 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 567 words, total size 5 kb.

Monday, May 29

Geek

Daily News Stuff 29 May 2023

Automated Poop Emojis Edition

Top Story


Tech News

  • The Snapdragon 8cx gen 4 is 32% faster than Apple's M4, at least on one benchmark.  (WCCFTech)

    Of course, neither chip is out yet, so take that with a sack of salt.

    But it does look like the commodity Arm market is finally shaking off the dust and competing seriously with Intel and AMD for laptop parts.


  • Meanwhile Arm announced an entire family of new cores, with 40% better performance, making them 15% faster.  (AnandTech)

    The catch there is that mobile chips come with a mix of fast, power-hungry cores, and slower more efficient cores.  It's the slower cores that are 40% faster, and the faster cores are only 15% faster.  That fast core - the Cortex X4 - is alternatively 40% more efficient for the same performance, but expect phone manufacturers to take the 15% and sacrifice your battery.

    But the mid-tier and low-end A720 and A520 cores are both faster and more efficient, so if you're in the market for a $500 phone - or a $250 one - rather than a $1000 model, things look a bit brighter.


Disclaimer: Ceci n'est pas une cuillère

Posted by: Pixy Misa at 05:52 PM | Comments (1) | Add Comment | Trackbacks (Suck)
Post contains 245 words, total size 2 kb.

Sunday, May 28

Geek

Daily News Stuff 28 May 2023

Frostbite Falls Edition

Top Story

  • Elon Musk has withdrawn Twitter from the European Union's "voluntary" program to control online "disinformation". (Tech Crunch)

    The media, and the bureaucrats, who believe they have the sole right to spread disinformation, are taking this about about as well as can be expected, with Thierry Breton, head of the European Union's Josef Goebbels Memorial Happy Fun Time Administration, issuing unveiled threats.

    What they're really upset about is that Twitter isn't playing the game:
    The pan-EU law, which entered into force back in November, requires VLOPs like Twitter to assess and mitigate systemic risks to civic discourse and electoral processes, such as disinformation.

    The deadline for VLOPs’ compliance with obligations in the DSA is three months from now.

    A request for comment emailed to Twitter’s press office returned an automated reply containing a poop emoji.
    (My italics.)

    You don't have to follow the rules. You're entirely expected to cheat. But you have to play the game.


Tech News

  • ChatGPT, Ace Attorney. (Volokh Conspiracy)

    A lawyer needed to file a motion to dismiss. Being busy, lazy, or probably both, he assigned the actual work of drafting the motion to another lawyer at his firm.

    Likewise busy, lazy, or probably both, the second lawyer handed the task to ChatGPT.

    Being a pathological liar in a box, ChatGPT invented multiple entirely imaginary cases as precedent.

    The judge is not amused.

    This is just the start of it. Expect a lot more of this until OpenAI becomes a penny stock and everyone goes back to looking up stuff on paper.


  • The "Hot Pixel" attack can leak data from almost any CPU at a rate of 100mbps. (Tom's Hardware)

    Note the lower case m. We're talking millibits - so about one letter or digit per minute.

    And it only works under perfect conditions, and requires access to run arbitrary code on the machine in question, so it's very likely you have more pressing concerns.


  • Fancy a bit of light housekeeping? Here's your chance. (The Guardian)

    The US government is giving away excess lighthouses made redundant by GPS, though you have to be a federal, state, or local government entity or an approved registered non-profit to qualify for a freebie.

    If nobody on the list wants the lighthouses they will go to public auction.


  • Amazon office workers are planning to go on strike over, uh, over having to work in an office. (CNN)

    Thanks, said Amazon in a short note, though we weren't planning to announce the next round of layoffs just yet.


  • Where's that story where the paid staff at a mostly-volunteer help line unionised and were all fired and replaced with ChatGPT?

    Oh, here we go. (Gizmodo)

    Right under an ad proclaiming that ChatGPT is revolutionising customer service. Yes indeedy.

    You'd need to have a kidney of stone not to laugh.

    Attention kids: ChatGPT ain't gonna put plumbers out of work, and you won't have $160k in student debt.


  • Sales of the newly launched RTX 4060 Ti and Radeon 7600 graphics card are even more miserable than the miserable sales of other miserable graphics cards in this miserable generation. (Notebook Check)

    Part of the problem is inflation, particularly that official inflation figures are a lie, and that where gamers expect graphics cards to get cheaper with each passing year, for once the costs to manufacture the cards have increased sharply.

    And part of it is that Nvidia is coming off three years of government lockdowns, crypto mining crazes, and money printer go brrr where they could sell everything as fast as they could make it, and now that the economy has predictably gone splut nobody is in the mood anymore.

    Nvidia doesn't care because it's happily selling high end cards at the price of a new car to the tech scam du jour, which is to say, AI. 

    And all AMD needs to do is be slightly cheaper than Nvidia.  AMD created the chips for both the Xbox Series S / X and PlayStation 5 anyway, so they have that entire segment of the market locked up.

    Gamers for whom money is no object, and professionals for whom time is money, have already bought high-end cards. Those who need to watch their budgets are buying last year's models on clearance. Nvidia's RTX 3060 12GB model (not the cut-down 8GB model), and AMD's Radeon 6700, 6700 XT, and 6800 are all good options.

    And there's also Intel, which seems to have cleaned up its early driver mess, and is offering the Arc A750 at very attractive prices.  If Intel can just make a decent card at a decent price with its upcoming "Battlemage" and "Celestial" cards, it might stand a chance of gaining significant marketshare.  But those aren't expected to start showing up until at least the end of the year, and more likely next year.



Disclaimer: There is no spoon. Currently unavailable. We don't know when or if this item will be back in stock.

Posted by: Pixy Misa at 04:45 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 830 words, total size 7 kb.

Saturday, May 27

Geek

Daily News Stuff 27 May 2023

Termites R Us Edition

Top Story



Tech News

  • Why have air conditioning when we could all live like termites?  (Ars Technica)

    Well, apart from the fact that you're talk about demolishing every human dwelling in the world and rebuilding from scratch, apart from the infrastructure problems of such dense living spaces - not to mention the psychological and sociological ones, apart from the fact that termite mounds are found in very few and specific locations in the world for a very good reason, apart from the fact that the one example presented is located in a city at an altitude of a mile where nobody needs AC in the first place, apart from the fact that in any cold climate the people on the outer edge of the building - you know, the only part that has windows - would all freeze to death when winter arrives - termite mounds not notably common in Norway, for example, apart from all that, the square-cube law would mean that twice a day you'd have hurricane-speed winds blasting through the insides of your incomprehensibly expensive new megastructures.

    Apart from that, sure.  Eat the bugs and live like them too.  I'll be over here.  In my house.  Eating steak.  Or chicken nuggets anyway, given the price of a good steak.


  • Electric truck maker Nikola is at risk of being deleted from NASDAQ mostly because it hasn't made any electric trucks.  (Tech Crunch)

    It has faked some videos of electric trucks, though, so maybe Disney will buy it.


  • IMEC - the global body that tries to manage the semiconductor industry the way an ailurophobe manages thirty to fifty feral cats - has laid out plans to reach the 2A node by 2036.  (Tom's Hardware)

    2A - two Angstroms - is 0.2 nanometres.  Which is slightly smaller than a single silicon atom.

    It might seem impossible to construct silicon chips with features smaller than silicon itself, and you'd be correct, except for the fact that these are not engineering numbers but marketing numbers, which is to say, lies.

    And that would mean the whole plan is nonsense except for the fact that the current mainstream production nodes - 7nm and 5nm - are also marketing numbers.

    So, yeah, chips are going to keep getting larger and smaller and more complicated at a rapid pace for at least another decade.


  • Am I the asshole unethical one?  (Daily Nous)

    An ethics professor - I've made my opinions on the field of ethics clear before - made it very clear to his class that if they cheated on their exam, they would fail.

    Then he posted a sample exam with obviously incorrect answers to a known cheating site.  We're talking about 2 + 2 = banana kind of answers.

    Then his students cheated.  

    At least he didn't have to fail them for cheating, because they failed in the old-fashioned way of getting zero on the test.

    I think they have a bright future as ethicists.


  • Don't buy HP printers.  (The Verge)

    HP offers a "Plus" program where - if you sign up within seven days of buying your new printer - you get "free" ink for "six months".

    Oh and also HP locks your printer to prevent you ever buying non-HP ink cartridges, even if you later cancel your subscription, even though HP inkjet printers are certified as not locking out third-party ink cartridges.

    Epson and Brother both sell inkjet printers with ink tanks that - this is complicated, so bear with me - you fill with ink.


  • "China's" "home-grown" "Powerstar" CPU is a painted-over Intel 10th generation Core i3.  (Tom's Hardware)

    Not in the sense that China stole Intel's design and made the chips itself while claiming credit for the design effort, since Intel's 10th generation chips were made on a 14nm process and China does have 14nm production capability, but in the sense that these are made by Intel and then literally painted over with new part numbers.

    Which is one way to do it, I guess.


  • Meanwhile in Real China TSMC is preparing a 6x reticle size CoWoS-L super carrier interposer for extreme SiP processors.  (AnandTech)

    You know how a few years ago AMD launched its new Zen CPUs, and rather than making an 8-core chip for desktops and a 16-core chip for workstations and a 32-core chip for servers (which they couldn't do because they didn't have any money), they made a single 8-core design that you could use one or two or four of according to your need?

    That's what this is about.  Only six times bigger.



Disclaimer: Nothing in the above content, or in the below content, if any, or in any other content to either side, in front of, behind, or in any other orthogonal or non-orthogonal dimension, at any distance, should be taken, construed, inferred, or assumed to be a statement of fact, opinion, or an orthographically, grammatically, or syntactically correct sentence in any language, extant, extinct, imaginary, or hypothetical.  There is a spoon.  Just $4.99, or $24.99 for a package of four.  What a deal.

Posted by: Pixy Misa at 05:16 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 957 words, total size 8 kb.

<< Page 1 of 613 >>
111kb generated in CPU 0.0311, elapsed 0.3822 seconds.
60 queries taking 0.3645 seconds, 367 records returned.
Powered by Minx 1.1.6c-pink.
Using https / https://ai.mee.nu / 365