Oddly when activating reader mode in the browser it contained text from a different article (apparently one that follows the linked article in truncated form). Can't recall the last time I've seen that happen.
2,552 karma · joined November 8, 2013
So it falls back to 'you're right', rather than be arrogant or try to save face by claiming it is correct. Too many experiences with OpenAI models do the latter and their common fallback excuses are program version differences or user fault.
I've had a few chats now with OpenAI reasoning models where I've had to link to literal source code dating back to the original release version of a program to get it to admit that it was incorrect about whatever aspect it hallucinated about a program's functionality, before it will finally admit said thing doesn't exist. Even then it will try and save face by not admitting direct fault.
Third-party tools have tried to reimplement it but it's either been by bastardizing the native W11 horizontal taskbar to be vertical (eg: Windhawk) or just restoring the old W10 taskbar code (eg: StartAllBack).
In past HN comments this apparently exists IRL in Germany and/or Canada, where age can be proven via a smartphone without leaking one's identity to the verifier and without any communication back to the government.
While there's much talk about uBlock Origin with Mv2 other losses include the last remaining Javascript managers for Chromium like ScriptSafe that have no Mv3 counterpart.
Are any of the bug reporters using Vivaldi? As for some reason that browser allows (and enables by default) clearing extension storage when clearing history/cookies via Delete Browsing Data.
This is something the Stylus addon dev noted when a user was reporting the addon wasn't remembering their settings and it was because the user had unwittingly wiped the extension storage due to the browser defaults.
(I actually use Vivaldi myself but virtually never use that feature so was unaware of the behavior until reading about the bug report)
That said, they're all GIFs and each ~10-22MB. Making loading the readme larger than the program size itself. Embedding some video would be snappier.
There's a Unicode sequence that tries to use a monochrome glyph instead if it's supported which I prefer as it's more in keeping with the rest of the text (though an issue with some of those variants is legibility at small sizes/PPI).
ZIP similarly doesn't stand for anything but is styled as such. So I suppose the pushback is more against being used as a stand-in for generic raw image formats.
Edit: just noticed your comment was greyed but I didn't downvote it fwiw.
It's interesting as when I last looked a couple years ago only Fujifilm had a mirrorless digital camera (with cropped sensor) that supported this on-device.
Otherwise it seemed like the most promising alternative I've seen that satisfied various requirements.
The link you cite though was careful to avoid making claims that couldn't be substantiated. It lists only what is in the database locally and the telemetry section doesn't include image content/metadata but user interactions with the app itself.
I've seen something similar occur for some popular Youtube videos, too. A video author will fire up some arbitrary Windows setup, which can come bundled with third-party software and use Bing for various things including weather in the taskbar and queries in the search bar, then open Wireshark to scaremonger with DNS queries, accusing Microsoft of spying just for requests made by the services/programs/features they have enabled in their install.
When often cursory lookups of the domains in search engines show what their purpose is and are contrary to such videos' alleged (and worse, guessed) purpose.
It's a problem as there are legitimate concerns with certain aspects of Windows software with non-privacy respecting defaults but for an average user it gets muddled with irrelevant/incomplete info that doesn't lead to high quality actionable results.
It's interesting as typeface plagiarism became rampant beginning in the 70s when more accurate photo reproductions made it trivial. This was problematic for designers wanting to make a livelihood, which is something ITC sought to mitigate by better up-front payments (IIRC from U&lc's coverage) to incentivize quality typeface creation.
There's a distinction though between literal plagiarism and just inspiration from elements. US copyright law doesn't protect either for typeface forms but ironically it does allow copyright for the code used in font files.
I've seen OpenAI's o3-mini (their reasoning model) suggest verbatim code and comments that I found on Github predating LLMs by years. It seems the more times the same code and comment appears online the more likely this is to occur. I'd imagine there would be studies looking into the scope and frequency this occurs and how much is considered fair use.
In both IRL and online discussions sometimes a stranger is the perfect person to talk to about certain things as they have no history with you. In ideal conditions for this they have no greater context about who you are and what you've done which is a very freeing thing (can also be taken advantage of in bad faith).
Online and now LLMs add an extra freeing element, assuming anonymity: they have no prejudices about your appearance/age/abilities either.
Sometimes it's hard to talk about certain things when one feels that judgment is likely from another party. In that sense chatbots are being used as perfect strangers.
Edit: for clarity, I believe anything with the ability to analyze the user environment via Javascript/etc on major sites is likely fingerprinting regardless. Blocking, environment isolation and spoofing is already necessary to mitigate this.
Not that I wouldn't similarly expect it from Miyazaki in terms of general generative art but the actual source of all the articles/memes about his quote point to a 2016 video where he's being demo'd a disturbing 3D simulation of an oily looking human figure crawling on the ground by its head while the dev explains to Miyazaki and others that 'it feels no pain so it learned to move by its head' and it could be used for horror games.
It's then that Miyazaki expresses the 'insult to life itself' quote and explains the devs have no idea what human pain is. Makes one wonder how the devs thought the reaction would be any different tbh.
Edit: reading that he clarified in an interview[1] a couple years later that his distaste was due to believing the dev was aiming at humorizing such body contortions of realistic humans which he took issue with.
Consider for example the process of casually learning about a subject in a non-academic scenario. People frequently will search online, ask on forums/chat and watch videos. Search results are widely critiqued for being poor these days, so people often turn to asking communities for specific questions.
Sometimes the only communities available are beginner hostile, expecting some prerequisite understanding and even being antagonistic toward knowledge gaps (mostly because it's uninteresting to them and they've seen so many of such questions it introduces a jadedness).
It's scenarios like that where non-judgmental, always available LLMs stand out in contrast most. I have wondered though how that might shape the dependency on LLMs broadly if people get used to not having to deal with the rough edges of humans.
Rather than directly admit fault they'll regularly respond in subtle (moreso o3) to not so subtle roundabout ways that deflect blame rather than admit direct fault, even when it's an inarguable factual error about even conceptually non-heated things like API methods.
It's an annoying behavior of their models and in complete contrast to say Anthropic's Claude which ime will immediately and directly admit to things it had responded incorrectly about when the user mentions it (perhaps too eagerly).
I have wondered if this is something its learned based on training from places like Reddit, or if OpenAI deliberately taught it or instructed via system prompts to seem more infallible or if models like Claude were made to deliberately reduce that aspect.
Do you mean in terms of open source vector editors? As there a wide variety of tools with SVG authoring/editing capability, among the most well-known being Adobe Illustrator, Sketch, Affinity Photo/Designer, even some web apps are available that were made for online SVG editing (eg: SVGator).
Inkscape, like some tools such as Affinity's, adds its own XML namespace with custom attributes and values, though for arrows I would expect it to use native `marker` elements.
It's certainly true that with SVG's flexibility and particularly with cross-browser handling differences/bugs it can become its own task to get consistent presentation when doing more complex things with it. Still very fond of the format.
An unfortunate recent change for anyone using Amazon as a general review resource is they login-gated links to everything except the summary page. So without an account or logging in one can't view lower starred reviews.
The problem is too many search libraries on sites are anywhere from useless to completely broken, even for some household name brand sites. My go-to test is while on an existing product page to search for either some part of the product title or a snippet of bullet point feature verbatim, to see if the page I'm currently on gets returned at all. It's stunning how many sites fail this test.
Sony capitalized on the larger always online controversy with a low budget ad poking fun at the concept[1], which was great marketing.