• 10 Posts
  • 228 Comments
Joined 1 year ago
cake
Cake day: October 4th, 2023

help-circle



  • wordfreq is not just concerned with formal printed words. It collected more conversational language usage from two sources in particular: Twitter and Reddit.

    Now Twitter is gone anyway, its public APIs have shut down,

    Reddit also stopped providing public data archives, and now they sell their archives at a price that only OpenAI will pay.

    There’s still the Fediverse.

    I mean, that doesn’t solve the LLM pollution problem, but…



  • https://www.gov.uk/government/news/government-launches-crackdown-on-mobile-phones-in-schools

    Mobile phones are set to be prohibited in schools across England as part of the government’s plan to minimise disruption and improve behaviour in classrooms.

    New mobile phones in schools guidance issued today (19 February 2024) backs headteachers in prohibiting the use of mobile phones throughout the school day, including at break times.

    Many schools around the country are already prohibiting mobile phone use with great results. This guidance will ensure there is a consistent approach across all schools.

    I suppose if enough countries do that sort of thing, pagers might start doing a comeback.

    EDIT: Though looking at the wording, I’m not actually sure if this is a “we’re banning cell phones” or a “we’re talking about policies that make it look like we’re banning cell phones to keep the anti-cell-phone crowd happy”.


  • I have not used it, but labwc is apparently an openbox-alike compositor for Wayland.

    Firstly, we believe that there is a need for a simple Wayland window-stacking compositor which strikes a balance between minimalism and bloat approximately at the level where Window Managers like Openbox reside in the X11 domain. Most of the core developers are accustomed to low resource Desktop Environments such as Mate/XFCE or standalone Window Managers such as Openbox under X11. Labwc aims to make a similar setup possible under Wayland, with small and independent components rather than a large, integrated software eco-system.


  • There are a large number of people in Hezbollah. Israel is fighting them.

    You’re talking about using a Hellfire R-9X.

    In order to launch those concurrently against, I dunno, sounds like there are maybe hundreds or thousands of targets, you’re going to need to have hundreds or thousands of drones. You’re gonna need something like a TB-2 at least to be lobbing them, not a tiny little drone. You’re talking about a lot of medium-size UAVs. That’s where your scale limitation is gonna come from.

    Those things are fine if you’re trying to kill one person. But Israel’s fighting a number of people, even if it can identify them. They aren’t gonna have thousands of drones above Lebanon.

    And if they’re hitting buildings and such, then you’re gonna be collapsing buildings and stuff like that.

    Secondly, I assume that the Lebanese government is not going to give Israel free reign to do drone strikes on Hezbollah on Lebanese territory, will shoot at those drones, so to use those, you’d need to destroy any air defense that Lebanon has. My guess is that Israel’s looking to just fight Hezbollah as much as possible.



  • There’s still gonna be some collateral damage with those, that can’t be employed at scale as readily – you’d have to concurrently target huge numbers of people from airborne platforms, and these are pretty small charges. Given that Hezbollah isn’t fighting in the open – understandably – this is probably about as good as it realistically gets in terms of collateral damage.

    Israel could maybe use DIME charges to have a smaller difference between lethal radius and damaging radius, but that’s got its own unpleasant aspects.

    https://en.wikipedia.org/wiki/Dense_inert_metal_explosive

    Dense inert metal explosive (DIME) is an experimental type of explosive that has a relatively small but effective blast radius. It is manufactured by producing a homogeneous mixture of an explosive material (such as phlegmatized HMX or RDX) and small particles of a chemically inert material such as tungsten. It is intended to limit the effective distance of the explosion, to avoid collateral damage in warfare.

    Upon detonation of the explosive, the casing disintegrates into extremely small particles, as opposed to larger pieces of shrapnel which results from the fragmentation of a metal shell casing. The HMTA powder acts like micro-shrapnel which is very lethal at close range (about 4 m or 13 ft), but loses momentum very quickly due to air resistance, coming to a halt within approximately 40 times the diameter of the charge. This increases the probability of killing people within a few meters of the explosion while reducing the probability of causing death and injuries or damage farther away. Survivors close to the lethal zone may still have their limbs amputated by the HMTA microshrapnel, which can slice through soft tissue and bone.

    If Israel isn’t using those already, I guess we could send 'em some, if we have some sitting around. Realistically, though, I doubt that collateral damage is gonna be possible to reduce a whole lot, given the fact that Hezbollah’s hiding in a civilian population.



  • looks dubious

    The problem here is that if this is unreliable – and I’m skeptical that Google can produce a system that will work across-the-board – then you have a synthesized image that now has Google attesting to be non-synthetic.

    Maybe they can make it clear that this is a best-effort system, and that they only will flag some of them.

    There are a limited number of ways that I’m aware of to detect whether an image is edited.

    • If the image has been previously compressed via lossy compression, there are ways to modify the image to make the difference in artifacts in different points of the image more visible, or – I’m sure – statistically look for such artifacts.

    • If an image has been previously indexed by something like Google Images and Google has an index sufficient to permit Google to do fuzzy search for portions of the image, then they can identify an edited image because they can find the original.

    • It’s possible to try to identify light sources based on shading and specular in an image, and try to find points of the image that don’t match. There are complexities to this; for example, a surface might simply be shaded in such a way that it looks like light is shining on it, like if you have a realistic poster on a wall. For generation rather than photomanipulation, better generative AI systems will also probably tend to make this go away as they improve; it’s a flaw in the image.

    But none of these is a surefire mechanism.

    For AI-generated images, my guess is that there are some other routes.

    • Some images are going to have metadata attached. That’s trivial to strip, so not very good if someone is actually trying to fool people.

    • Maybe some generative AIs will try doing digital watermarks. I’m not very bullish on this approach. It’s a little harder to remove, but invariably, any kind of lossy compression is at odds with watermarks that aren’t very visible. As lossy compression gets better, it either automatically tends to strip watermarks – because lossy compression tries to remove data that doesn’t noticeably alter an image, and watermarks rely on hiding data there – or watermarks have to visibly alter the image. And that’s before people actively developing tools to strip them. And you’re never gonna get all the generative AIs out there adding digital watermarks.

    • I don’t know what the right terminology is, but my guess is that latent diffusion models try to approach a minimum error for some model during the iteration process. If you have a copy of the model used to generate the image, you can probably measure the error from what the model would predict – basically, how much one iteration would change an image or part of it. I’d guess that that only works well if you have a copy of the model in question or a model similar to it.

    I don’t think that any of those are likely surefire mechanisms either.







  • That characteristic sound you hear associated with falling bombs in movies and such was originally a noisemaker on World War II German dive bombers and bombs intended to intimidate soldiers on the receiving end.

    https://www.slashgear.com/1370552/stuka-siren-ju-87-noise-explained/

    As well as saving the Stuka from early retirement, it is thought that Ernst Udet also suggested its most famous feature — the siren (some sources say that this was an intervention by Hitler himself). The sirens were fitted to the legs of the plane’s fixed undercarriage. They were driven by propellers that spun in the airflow, and could be activated and deactivated from the cockpit.

    The psychological effect of the siren was best explained French general Edouard Ruby, who reportedly said that on hearing the terrifying wail, his infantrymen “cowered in the trenches, dazed by the crash of bombs and the shriek of the dive bombers.” But many Stuka pilots also didn’t like them. The sound was just as audible in the cockpit of the Stuka as it was to forces on the ground, and the bulky sirens added weight and reduced the speed of the already slow bomber. Reportedly, some squadrons fitted simple air whistles to the Stuka’s bombs instead, creating the famous “falling bomb whistle” that Hollywood still insists that all ordnance makes as it plummets to earth.

    Both the Stuka’s terrifying wail and the falling bomb whistle became so famous that they have since become standard stock sound effects in movies, used whenever any airplane dives at high speed or any bomb is dropped. But, unless you’re old enough to have been on the battlefields of Europe in the very earliest days of WWII, these are sounds that you’ll only ever hear in movies now.

    Stuka dive sound:

    https://www.youtube.com/watch?v=UQzv-8pJSqY

    Falling bomb whistle:

    https://www.youtube.com/watch?v=zlsHYKkmHoI

    Maybe one could do the same thing with grenades.



  • Well, they did remove it when they found out. But…

    Look. I’m looking at a Thinkpad. Lenovo owns that line now. I dunno if they can push firmware updates to old, pre-Lenovo models, but they can to current versions. Those things are pretty common in a business setting. AFAIK, the US has never raised any issues with Lenovo and security a la Huawei. But if there was an honest-to-God, knock-down, drag-out war, I assume that Beijing is gonna see whether it can leverage anything like that. And I’ve got, what…a microphone? A camera? Network access? Maybe interesting credentials or other things in memory or on my drive? I mean, there are probably things that you could do with that.

    Then think of all the personal phones that military people have. Microphone. Camera. Network access and radio. Big fat firmware layer.

    My guess is that if you did a really serious audit of even pretty secure environments, you’d find a lot of stuff floating around that’s potentially exploitable, just due to firmware updates. If you exclude firmware updates, then you’re vulnerable to holes that haven’t been patched.

    Okay, maybe, for some countries, you can use all domestic manufacturers. I don’t think that South Korea could do that. Maybe the US or China could. But even there, I bet that there are supply chain attacks. I was reading a while back about some guy selling counterfeit Cisco hardware. He set up a bunch of bogus vendors on Amazon. His stuff got into even distribution channels with authorized Cisco partners, made it into US military networks.

    https://arstechnica.com/information-technology/2024/05/counterfeit-cisco-gear-ended-up-in-us-military-bases-used-in-combat-operations/

    Counterfeit Cisco gear ended up in US military bases, used in combat operations

    That guy was just trying to make a buck, though I dunno if I’d have trusted his products. But you gotta figure that if that could have happened, there’s room for intelligence agencies to make moves in that space. And that’s the US, which I bet is probably the country most-able to avoid that. Imagine if you’re a much smaller country, need to pull product from somewhere abroad.