@miki@dragonscave.space avatar

miki

@miki@dragonscave.space

blind coder / comp-sci student, working in automatic speech recognition for CLARIN. Polish. Libertarian leaning. Feel free to get in touch.

This profile is from a federated server and may be incomplete. Browse more on the original instance.

miki, to random
@miki@dragonscave.space avatar

Am I the only person who gets surprised when a C/C++ project actually builds successfully on first try?

miki, to random
@miki@dragonscave.space avatar

This whole Microsoft Recall thing makes me want to return to my "permanent storage of speech history" idea. Annotate it with some metadata like timestamps, app name and window title, stick it in a vector database for RAG, and some really interesting possibilities start to emerge.

miki, to random
@miki@dragonscave.space avatar

If you truly want to be inclusive to blind people in meetings, virtual or otherwise, forget describing your appearance, just make sure you don't nod, shake your head or point a finger without verbally describing what you're doing. I literally just got bitten by this exact thing.

miki, to random
@miki@dragonscave.space avatar

Google Drive for Mac went from hands-down the best cloud app I've ever seen on any platform to an utter piece of garbage.

miki, to random
@miki@dragonscave.space avatar

I think we severely underappreciate the impact of IP address exhaustion on basically everything on the modern internet.

If not for that issue, dynamic IP addresses and NATs wouldn't have to exist, and that would make P2P-based protocols a lot more practical.

miki, to random
@miki@dragonscave.space avatar

It took Python over 30 years to make "exit" in the terminal actually exit instead of telling you what to do to exit.

I know why this works the way it works (the message is actually the string representation of the exit function, and making the string representation of "exit" quit the interpreter would be just as counterintuitive), but this ugly hack should have been fixed in like 0.2 for God's sake.

miki, to random
@miki@dragonscave.space avatar

There's a fundamental conflict of interest between accessibility experts and accessibility overlay vendors. If accessibility overlays ever get good enough to replace the experts, which, just to be clear, they haven't yet, they'll put the experts out of their jobs. I wouldn't be surprised if accessibility "experts" keep spreading FUD about overlays long past the point when they actually become useful. After all, making themselves sound irreplaceable might be the only viable strategy for them.

This is just something to ponder when hearing what experts say about AI in accessibility.

miki, to random
@miki@dragonscave.space avatar

The fundamental problem with the concept of alt text is that it's language specific, while pictures are not.

There are plenty of social media accounts that post pictures and videos that are universally funny, beautiful or interesting, regardless of the language you speak.

This is one more reason why fixing accessibility problems by inventing better technology for ourselves is a better approach than fixing accessibility by advocating for it. Even if you teach every single content creator on the planet to write alt descriptions and write them well, which is a monumental task in itself, this still won't solve the problem for the creators that have a multilingual audience. If you teach an AI to write good alt descriptions for those pictures, the problem will be solved for everyone.

To be clear, we're not quite at that point yet, as good as GPT Vision now is, it still doesn't beat a skilled human describer, but I think we'll get there quite soon.

miki, to random
@miki@dragonscave.space avatar

There's "AI" in "braille"

miki, to random
@miki@dragonscave.space avatar

IMO third-party apps not having access to system APIs is a far bigger antitrust issue than petty squablles between big American companies and big European companies about who deserves what share of the pie.

AI companies would have a serious chance of disrupting the Apple / Google duopoly, but that's not possible if you can't make an AI assistant app that responds to a wake word, can make calls without the user having to touch the screen, can read and respond to texts, access call audio for transcripts / summaries etc.

Android is slightly better than iOS here, but only slightly.

Alternative App Stores or even unrestricted sideloading don't solve this, Android has the latter, but there's still no way for apps to accomplish many of these things without an exploit that lets you root your device.

miki, to random
@miki@dragonscave.space avatar

What's up with this weird connection between American colleges and protests?

This seems to be an exclusively American thing. We've had student protests in the past, but they were about extremely serious issues happening here and affecting those studennts' daily lives, not every single issue that is popular and trendy at the time, from the Vietnam War to BLM to Gaza.

This seems to be an exclusively AMerican phenomenon.

miki, to random
@miki@dragonscave.space avatar

Things you wouldn't predict 20 years ago.

Being very knowledgeable about transformers can easily net you a seven-figure salary.

miki, to random
@miki@dragonscave.space avatar

I've just seen a commandline webserver that pops up a "enter your email if you want to sign up to our newsletter" prompt on first run. The world we live in.

miki, to random
@miki@dragonscave.space avatar

Screen reader tip:

If you often have long-running operations in your terminal, don't want your screen reader to talk constantly while they're going but want to be notified when they finish, define the following alias (works for ZSH and Bash):

beep() {
$@
echo -e '\a'
}

And then execute your commands like:

beep python long_running_script.py

You can set your screen reader to quiet, and your terminal will beep once the command is done.

I also have a more advanced version of this that sends any error output to my phone via Telegram in case of failure.

miki, to random
@miki@dragonscave.space avatar

One realization I recently had about AI:

None of the instruction-tuned models I'm aware of have really been trained to ask clarifying questions.

Answering a computer question makes no sense if you don't know what operating system the user is on. Answering a car question makes no sense if you don't know what kind of car it is. Answering a legal question makes no sense if you don't know what jurisdiction.

If you call an expert and ask them a generic question like "what to do if there's nothing on my screen", the expert will have a conversation with you. Before trying to answer, they'll ask what kind of computer it is, Pc or Mac, desktop or laptop, what kind of troubleshooting was performed, whether you can hear the fan going etc. LLMs generally don't do this, which makes them answer in the most useless and generic way possible.

I wonder whether there's something that makes it much harder to fine-tune LLMs on longer conversations, or whether nobody has just tried this so far.

miki, to random
@miki@dragonscave.space avatar

We've reached the point where when I disagree with an LLM, the LLM is right more often than I am. Now who's the hallucinator in this relationship?

miki, to random
@miki@dragonscave.space avatar

Hypothetical question. How would the world of 2024 look like if Jobs's famous 2007 iPhone demo failed?

For context, the iPhone software was notoriously unstable at the time, Jobs was given a list of actions he was supposed to follow, in a very specific order, which was, in theory, supposed to work. It did actually ended up working, but there was a very high chance of the whole thing going awry.

The iPhone completely revolutionized the phone market, in a way most people don't even appreciate. Multitouch screens and fancy animations were one thing, but what matters a lot more IMO is what it did to carriers. Most if not all smartphones that came before were deeply tied to their networks. Direct, computer-style TCP/IP communication with unaffiliated servers on the internet was rare. For example, BlackBerry used custom protocols and needed RIM hardware in your carriers datacenter, Europe was big on WAP, which was a weird binary version of HTTP that your carrier's gateway would translate to normal requests, while Japan was big on iMode, which essentially was its own walled garden which had little to do wwith the "big" internet. The ability to use a phone as essentially a small computer, using your carrier's network instead of WiFi, was pretty revolutionary.

A lot of people complain about Apple's ecosystem being a walled garden, but I'd rather have a somewhat developer-friendly, global, unified system under Apple than the fragmented, country-specific, carrier-specific mess that existed before.

miki, to random
@miki@dragonscave.space avatar

Today's code find of the day:

A Python global variable being stored in a single-element array so that it can be modified from inside a function.

Somebody just... didn't know how to use Global I guess?

miki, to random
@miki@dragonscave.space avatar

Is there a culture that celebrates new year's eve on april 13th, 4 minutes before 10 PM? I think so, and I think my neighbors come from that culture because I just heard a big round of fireworks being fired.

miki, to random
@miki@dragonscave.space avatar

In the last few months, we went from AI music being a curiosity you needed a good Linux box to play with, to something pretty crappy but looking somewhat interesting, to... this.

miki, to random
@miki@dragonscave.space avatar

I find hot-takes and speculations on the future from years ago to b just needlessly fascinating. Like this post from 2003, which discusses whether there's even a point to "moblogging" (AKA blogging on your phone).

https://www.russellbeattie.com/blog/1001648

miki, to random
@miki@dragonscave.space avatar

Unpopular opinion.

"maybe later" buttons are kind of nice, actually.

If you have an actual "no" button, you need to figure out what to do when the user accidentally clicks "no" and wants to change their decision.

You either need a setting for every possible pop up you might want to display, or at least a "reset warnings" button somewhere. That's a lot of maintenance, not to mention the fact that most users don't know how to find those.

miki, to random
@miki@dragonscave.space avatar

Jesus, Google drive is slow

miki, to random
@miki@dragonscave.space avatar

Is anybody seriously using Beeware Toga, particularly in an accessibility context? It seems like one of the most accessible cross-platform GUI frameworks out there, but the lack of popularity gives me pause.

miki, (edited ) to random
@miki@dragonscave.space avatar

Voice Dream's upcoming changes are violating Apple's App Review guidelines, specifically guideline 3.1.2(a). If this concerns you and you don't want these changes to actually be made, I suggest doing the following:

  1. If you have friends at Apple, particularly in Corporate, ask them to find somebody to report the violation to, probably at the App Review or accessibility teams.

  2. send an email to Apple and tell them that this is happening. Be professional and not angry, use a corporate email address if you can, name.surname@provider if you cannot. Have a signature with your postal address and phone number, look like a professional, not a "random blindie". Mention that they're taking away features from users who already paid and mention the specific guideline number. You should also emphasize how important the app is to you and to the disability community in general. Bring their attention to the fact that the app has an Apple design award and is featured in App Store collections. Tell them why built-in iOS features aren't enough and how the app changes your life. Send the email to accessibility@apple.com, abuse@apple.com, abuse@icloud.com and tcook@apple.com. Those are not the right people to handle this, but the people manning[*] these email addresses are very likely to have far more agency than ordinary customer support representatives, and there's no way to contact the relevant people directly. If we cause enough of a ruckus, somebody is going to notice and get the message across to the right department, and you don't just ignore messages from the office of the CEO, even if they don't actually come from the CEO directly.

Edit: I originally mistyped the guideline number.
To be continued.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • JUstTest
  • everett
  • DreamBathrooms
  • khanakhh
  • magazineikmin
  • tacticalgear
  • Youngstown
  • osvaldo12
  • slotface
  • ngwrru68w68
  • rosin
  • thenastyranch
  • kavyap
  • InstantRegret
  • provamag3
  • Durango
  • mdbf
  • tester
  • cisconetworking
  • ethstaker
  • modclub
  • GTA5RPClips
  • cubers
  • Leos
  • normalnudes
  • megavids
  • anitta
  • lostlight
  • All magazines