One of the curl test servers (for SOCKS) had a help text output listing around ten command line options. One of the options it listed was never implemented and thus didn't work. The AI found out and reported.
(See how I cleverly did not mention AI in the title!) You know we have seen more than our fair share of slop reports sent to the curl project so it seems only fair that I also write something about the state of AI when we get to enjoy some positive a…
very nice post, what an adventure! I liked the note on ethics, it is a troubling problem indeed, and complex. Btw, I don't think I could manage receiving 200 bug reports in a week 😂
@pablo_martan Yes, he discussed slob, but in all the cases it was purely machine fabricated and not checked by a human expert. And I think that's what it is all about. If there's a human in the loop, it can be a handy tool. On a side note, the European AI Act also requires humans in the loop for the final check. And back in the days when there were large hopes for decision support systems, people also said they don't work without the humans in the loop. @bagder
@devSJR @pablo_martan I think even more importantly, this new kind of analyzer/AI tools are written for this purpose. This is not Joe Random asking chatgpt questions.
I've been thinking about your post a lot, especially after seeing such tools at my $dayjob. I'm biased due to their ethical impact, but even without it I consider them, on average, harmful. I write code, I make sure it works for my usage, I write tests, I do linters and static analysis, I do a peer review to share the knowledge and get external inputs. And then this thing, supposedly state of the art, goes over my code, mansplains it to me and finds either a false-positive (I wonder who removed false-positives from the lists you've got?), or a nit ("don't forget to add an extra check here!", "The comment is stale!"), or a misguided optimisation possibly introducing new bugs. I spend lots of time thinking over those useless blanket reports that ultimately don't matter because I have empirical evidence that my code works for my use case.
I have seen so called AI tooling generating "helpful reports", but they cannot replace decent tooling and tests. And yet some people replace their LSPs with LLMs :/
@nina_kali_nina all analyzer tools, including compilers, give a certain amount of false positives. I don't think we should expect AI tools to be any different. As long as frequency is manageable and there are decent ways to inhibit them.
The AI tools I've mentioned recently don't seem to have much more false positives than the state of the art static code analyzers we use also do.
I use Ultrasonic which streams from the with the "Music" app on my Nextcloud server. The Nextcloud music app supports both Subsonic and Ampache protocols.
None. I don't use any streaming junk. My music is on vinyls, 8-tracks, tapes, bands, CDs and similar things instead of being online. Yes, it may be old-fashioned and at times less practical, but my music played that way has something that the online "high-quality-umpa-umpa" lacks: soul.
Asking in English in hope that it'll reach more people! I'm french, windows user and unfortunately not playing locally. Do you know of some software a deafie could use to game with her pals? One that could transcript what people say in their mics? Not necessarily free, I'm willing to pay for something that works well.
Asking for a me.
Please, boost so a girl can play with her friends 🥰
@Poruri Sai Rahul from @FOSS United echoed the federation aspect of #XMPP (mentioning how he can communicate with Prav users using Quicksy) in an Indian Express article (indianexpress.com/article/opin…) in which he talked about how moving from one walled garden to another (WhatsApp to Arattai) is a bad move.
The author mentions being able to chat to telegram users from Matrix. He does not mention that: - this is not because Telegram is better than WhatsApp (or just marginally, with regards to alternative client tolerance) - this is possible via a matrix bridge - bridging to WhatsApp is also possible - XMPP bridges to telegram and WhatsApp exist too: slidge.im (yes, a shameless self-plug)
The word 'Vrede' jumped out at me from this 'Peace' installation. 'Vrede' is Danish for anger, fury, wrath. I wondered if it was an artistic provocation. But it seemed too confined to chance, that someone who happens to understand Danish happens to see this German artwork. So I looked it up and learned that 'vrede' is Dutch for 'peace'.
Vrede. Peace in Dutch. Wrath in Danish. I wonder if there's a word for words like these, that mean the opposite in different languages.
Krásné dobré ráno mastodonní bando! 🙋♀️🐈🐶 Zítra to bude rok, co odešel taťka. Mamka na dnešek naplánovala výlet na Portáš v Javorníku. Každoročně tam spolu jezdívali na borůvkové knedlíky. Pojede nás hodně. Jedna ze sester bere i tchána s tchýní. Docela nevím, co mám čekat. Prochajdu s Baldínem musím odložit. Vyrážíme ráno, protože sraz je u mamky. No, myslete na mě. Vy si užijte krásnou neděli plnou pohody, dobrot a nicnedělání! 😊 #dobre_rano #dobréRáno
In bizarre move, Framework embraces deeply extremist views
Framework, the maker of repairable laptops, is embroiled in a controversy, as the company and its CEO are openly supporting people with, well, questionable views.
If you know a little bit about PR in social media space, you might note that, right out of the gate, a project by a vocal white nationalist known for spli
I remember coming home from school and telling my mom Columbus "discovered" America. She told me it was a load of crap, and set me straight. Mom was a hippie who challenged the status quo. At the time, it embarrassed me, I'd get frustrated and just wanted her to be like the other parents. I realized later how lucky I was, and do the same with my daughter. #indigenous #fuckcolumbus #indigenouspeoplesday
On my AMD Ryzen 7 8845HS mini PC, NVDA is a bit sluggish in some cases in Firefox; e.g. cursoring through messages in Gmail folders. For reasons I don't fully understand, setting the processor affinity to a single CPU core and setting the process priority to "above normal" helps significantly, even when the CPU is nearly idle. I don't currently have the time/energy to debug the root cause for this or write a proper add-on, but I wrote an NVDA global plugin to make the change for me automatically when NVDA starts. If it breaks something, you get to keep all the pieces. ``` import ctypes
import globalPluginHandler
class GlobalPlugin(globalPluginHandler.GlobalPlugin): def __init__(self, *args, **kwargs): super().__init__(*args, **kwargs) p = ctypes.c_void_p(ctypes.windll.kernel32.GetCurrentProcess()) ctypes.windll.kernel32.SetProcessAffinityMask(p, ctypes.c_void_p(1)) ctypes.windll.kernel32.SetPriorityClass(p, ctypes.wintypes.DWORD(0x00008000)) ``` #nvdasr
@kev This isn't about changing overall CPU settings. It's about changing the way Windows schedules the NVDA process, both which core it uses and what priority it has relative to other running apps. You can do this from the Task Manager context menu, but the global plugin I provided does this automatically from within NVDA.
@kev I can understand above normal priority anyway but how does the afinity help? That's strange. I thought NVDA would have a couple relatively busy threads running at once and so should be distributed over a couple cores, especially when doing all that IPC and managing a virtual buffer.
@x0 @kev I honestly don't even really understand priority. If the CPU is near idle, there shouldn't be much contention (and thus nothing should have to wait), especially with multiple cores, since there are more than enough cores for all the things. As for affinity, NVDA is largely single core, primarily because GUIs (and thus accessibility APIs) are pretty much main thread only. Most of the work of a virtual buffer actually happens in NVDA's injected code in the GUI thread of the remote process; e.g. inside Firefox. I haven't tested this, but I think this is more likely to be the many cross-process COM calls NVDA makes to query the focus, etc., not the buffer stuff. That involves a lot of context switching.
@x0 @kev not sure about how NVDA manages its work, but restricting processor affinity can help keep memory cached closer to the core that's running the thread. if there's lots of state (either lots of small allocations or a few big ones), you can pay a price waiting for it all to be paged in
@vt52 That's my best guess too, but I'm just surprised by how much that seems to impact things in this case. There are a lot of very small IPC calls, which I guess could result in a lot of small allocations, but we're talking maybe 10-100, not thousands. @x0 @kev
@vt52 @x0 @kev It'd also be interesting to know why this problem anecdotally appears to be worse on AMD CPUs than Intel ones, though of course there are just so, so many variables there, no pun intended.
@kev My original post included the code. That can go in a .py file inside %appdata%\nvda\scratchpad\globalPlugins, but that requires developer scratchpad to be enabled. If that's not something you're familiar with, you're probably best leaving this one alone for now. I don't want to go through the process of creating this as an actual installable add-on at this stage.
Also, for some reason, the process priority does not get set with this code but the affinity does. However, if I run the code as a one-liner in NVDA's Python console, it returns "1" and does what it's suppose to.
@twynn Hmm, I can't repro that here. Both get set by the plugin. I'm running latest alpha, which... might make a difference because of 64 bit? I tried to write the code to be independent of that though.
Not sure if I'm checking it properly, but when I invoke the code in the Python Console, it returns 1, and Task Manager reflects the change. If I just, say, restart NVDA, only the affinity gets reflected. I don't need it anyway, so it's not a big deal.
I don't use email in a browser like that, so maybe that's why, but on my 2012 3570K, NVDA is never the thing that lags, it's always everything else around it. To be fair it's actually not a laggy machine at all despite it's well-warn age, but if anything does, it's never the screen-reader.
@FreakyFwoof What's really intriguing about my situation is that clearly, it's not really anything in particular that is slow, but rather, how NVDA gets scheduled by Windows. That's not a bottleneck I ever expected to be noteworthy. One of these days, I might try to profile it and figure out what's really going on, but definitely not today and probably not any time soon.
oh yeah I've been doing this for quite a while now on a HX370. I have very little experience with NVDA addons so entirely possible I messed something up somewhere, but it works for me. This let's you set CPU affinity and priority in NVDA settings. Addon: iamtalon.me/cpuPriority-1.0.0.… Code: code.iamtalon.me/Talon/cpu-aff…
@talon Even though I have no problems, I'm gonna try it anyway because I'm curious if I'm just used to this old machine being slightly laggy, and since I don't know any different, I just deal with it. Got me curious.
@FreakyFwoof If you don't have a cpu with high performance and high efficiency cores this will probably not do much for you. But hey been surprised before. Let me know what happens
@FreakyFwoof You might get a very very negligible boost from setting the process priority to high but honestly I'd be surprised if anything at all happened. lol
@talon I think you're right, honestly. I won't see any performance gains until I upgrade this machine's well, everything. It's running better than it ever has though, Jake hooked me up with a GPU instead of Intel onboard graphics, and I upped it from 16 to 32 GB RAM. Windows 10 but it's not end-of-life because of the I guess, EU mandate so I still get updates. Never felt so good lol
@talon GTX 970 for the curious. On a hunt for the 3770K from Ebay just to really max it out. Plan is to donate it to the boy once I'm officially done with it, would make a perfectly usable NAS/Linux box, refuse to scrap it just because MS says I should.
@FreakyFwoof That's what we've been doing as well. Our homelab is pretty much entirely made of old parts that eventually got upgraded. Very happy with this setup.
Oh wow. Surprised you need this on a CPU like that. Makes me wonder whether there is some weirdness with AMD processors that NVDA is hitting because as much as Intel irritates me for other reasons, I don't have this issue on my Intel laptop.
Not entirely sure, and also not even sure how I'd debug that. I'd imagine since it's a mobile CPU it's related to the power profiles, and changing them does improve things a little, but it makes a huge difference especially if you're running in a more energy saving mode. It switches to them especially when running on batteries and I haven't found out how I can prevent it from doing that, but to be honest sometimes I don't even want it to do that because it means giving up quite a lot of battery time. But setting CPU affinity and priority seems to do a good enough job that I kinda forgot about it until recently when I tried to get it working as an addon and not as a weird set of external scripts.
@talon My 8845HS is in a mini PC, so still a mobile-ish CPU but theoretically less of the power saving silliness. Obviously an older generation than yours though and I was thinking that was the cause, but I guess not. I'm definitely curious about the underlying cause, but now that I have an interim solution, I'm not quite curious enough to be motivated to dig into it further. I might get nerd sniped enough one day, we'll see.
yup, this is pretty much my situation as well. This works well enough for me that I don't feel like digging, and there are people out there that know much more about this than me that have a much higher chance of figuring out what's going on here.
As if today couldn't get any worse, the bill for the server that runs this instance came out today and it was £1.60 more than expected. I never got any emails saying I'd gone over any limits or anything, but anyway, I'm now £0.02 negative. Some support to help keep this server going and help my other services grow, as well as covering personal costs, would be greatly appreciated.
At a critical moment, the tech press are failing to connect the dots between Apple and Google's craven capitulations and the authoritarianism they have nurtured within their own ecosystems. Apple is now corrosive to democracy itself, and we have to get smarter about the way these forces interact:
Apple bent the knee for months, leaving many commentators to ask why. But the reasons are not mysterious: Apple wants things that only the government can provide, things that will defend and extend its power to extract rents, rather than innovate.
The majority of people hate rich people because it's been proven that many show narcissistic/psychopathic tendencies. You only have to look at the various big tech CEOs and many in the current administration to see this play out in real time.
@ZBennoui Yet I've met rich people and they are interesting people who don't even discuss money, of course, most were not billionaires, but millionaires, for sure.
Oh don't get me wrong, I'm not trying to imply that all wealthy people are narcissistic jerks who deserve nothing, but the fact of the matter is that many are. It also depends on how you got rich in the first place. If you were born into wealth, you will simply not relate to people who either don't have the extraordinary amount of money you do, or who have come into it honestly by doing good work. There's a lot of nuance here, as with everything.
@ZBennoui Most of the milionaires I've met, including even some blind ones, made businesses. I only met 1 person born into it, and he's british royalty.
When using Microsoft Word or Google Docs, don't just make text bigger and bolder to make it a heading. That will work for sighted users, but screen reader users will miss that and just hear it as normal paragraph text. Use actual heading styles, like level 1 through 6.
In addition to working better with screen readers, this also matters for Braille transcription (I used to work for a company that develops a Braille translator). If a document author properly uses styles, then the software can infer what she is trying to do and do an appropriate thing when formatting the corresponding Braille document. It is harder and a lot more error-prone for software to try to have logic to, say, try to infer that a heading is intended based on text being bigger and bolder.
In the late 1980's there was an MS-DOS program called PIANOMAN. Yup, in the docs, it's written in all caps like that. This program would let you compose, play, edit, and save tunes, and even compile them into self-contained executables to give to others who didn't have a copy of PIANOMAN themselves. This program's claim to fame was that it could create pseudo-polyphonic music through the monophonic PC speaker by playing different notes extremely rapidly. This was a shareware program (try it for a while, if you like it, send in a registration fee to get the latest version and another disk of sample songs). I just found the shareware distribution of version 4 on Archive.org. One of the included tunes, which exists here only as a compiled, self-contained executable, is a version of a Tom Lehrer classic. Naturally, I figured some folks might be interested in this. It's set up to play repeatedly until a key is pressed, so I've only included three playings in this recording, and the last one fades out. Enjoy! #TomLehrer
Ah, the memories! PIANOMAN was one of the programs I used to play with when my brother got his first PC back in the 80s. I never figured out how to record songs like those demos, but it was fun to play the keyboard like a monophonic piano. Drove everybody crazy when I played it during Computer class at school, but I didn't care.
If you put solar panels over all the fields where corn is grown for ethanol today, you could power all of North America and still harvest shade crops underneath them.
Anthropic stole the books I co-wrote in order to train their planet-burning hallucination machine, so I've filed a claim. They wept that if every plagiarised author does so, the company will go bankrupt. Boo emdash hoo. www.anthropiccopyrightsettlement.com
Pablo Martínez
in reply to daniel:// stenberg:// • • •daniel:// stenberg://
in reply to Pablo Martínez • • •A new breed of analyzers
daniel.haxx.seBubu reshared this.
Pablo Martínez
in reply to daniel:// stenberg:// • • •devSJR
in reply to Pablo Martínez • • •Yes, he discussed slob, but in all the cases it was purely machine fabricated and not checked by a human expert. And I think that's what it is all about. If there's a human in the loop, it can be a handy tool. On a side note, the European AI Act also requires humans in the loop for the final check. And back in the days when there were large hopes for decision support systems, people also said they don't work without the humans in the loop.
@bagder
daniel:// stenberg://
in reply to devSJR • • •Bradalot “”
in reply to daniel:// stenberg:// • • •@devSJR @pablo_martan
Another example of powerful tools used productively in the hands of experts (which famously create chaos in other contexts).
It's why "Don't try this at home." is a thing.
Osma A 🇫🇮🇺🇦
in reply to daniel:// stenberg:// • • •@bagder
Grampa
in reply to daniel:// stenberg:// • • •Nina Kalinina
in reply to daniel:// stenberg:// • • •I've been thinking about your post a lot, especially after seeing such tools at my $dayjob. I'm biased due to their ethical impact, but even without it I consider them, on average, harmful. I write code, I make sure it works for my usage, I write tests, I do linters and static analysis, I do a peer review to share the knowledge and get external inputs. And then this thing, supposedly state of the art, goes over my code, mansplains it to me and finds either a false-positive (I wonder who removed false-positives from the lists you've got?), or a nit ("don't forget to add an extra check here!", "The comment is stale!"), or a misguided optimisation possibly introducing new bugs. I spend lots of time thinking over those useless blanket reports that ultimately don't matter because I have empirical evidence that my code works for my use case.
I have seen so called AI tooling generating "helpful reports", but they cannot replace decent tooling and tests. And yet some people replace their LSPs with LLMs :/
daniel:// stenberg://
in reply to Nina Kalinina • • •@nina_kali_nina all analyzer tools, including compilers, give a certain amount of false positives. I don't think we should expect AI tools to be any different. As long as frequency is manageable and there are decent ways to inhibit them.
The AI tools I've mentioned recently don't seem to have much more false positives than the state of the art static code analyzers we use also do.